基于小波和支持向量机的多尺度时间序列预测  

Research on multi-scale prediction of time series based on wavelet and Support Vector Machines

作　　者：曲文龙[1] 李海燕[1] 刘永伟[1] 杨炳儒[2] 

机构地区：[1]石家庄经济学院计算机系,石家庄050031 [2]北京科技大学信息工程学院,北京100083

出　　处：《计算机工程与应用》2007年第29期182-185,共4页Computer Engineering and Applications

基　　金：北京市自然科学基金( the Natural Science Foundation of Beijing City of China under Grant No.4022008);河北省教育厅资助科研课题( the Research Project of Department of Education of Hebei Province; China under Grant No.Z2006313)

摘　　要：介绍了相空间重构和基于支持向量机的时间序列预测建模技术,提出了基于小波和支持向量机的复杂时间序列预测方法,利用小波对复杂时间序列进行多尺度分解,对重构后的近似序列和细节序列分别利用支持向量机进行回归预测并将结果融合。对股票数据进行预测,试验结果表明该方法预测精度高于单尺度支持向量机和神经网络预测方法,可用于复杂非平稳时间序列的预测。The technology of phase construction and modeling of time series prediction based on SVM（Support Vector Machines） was introduced firstly.A complicated time series predicting method based on Support Vector Machines and wavelet was proposed.h performances multiple-scaled decomposition on complicated time series using discrete wavelet transform.Then the reconstructed ap- proximate series and detail series were regressed and predicted respectively using SVM and the outcomes were composed finally. The prediction model was established and applied it to the stock data.Experimental result indicates that the prediction model has superiority over simple SVM and ANN（Artificial Neural Network） for it has higher prediction precision and is applicable to predicting complicated and unstable time series.

关 键 词：时间序列预测 小波 支持向量机 多尺度 数据挖掘 

分 类 号：TP311.13[自动化与计算机技术—计算机软件与理论]